(T1130-08-46) Process Development of API in Continuous Manufacturing

Purpose: When new products are being considered for continuous manufacturing, there are many critical steps needed to make the needed steps for production. This paper focuses on powder characterization and process development in a feeding-blending system and demonstrates the suitability of a continuous manufacturing route for an API. Due to the scarcity and potency of the API, a surrogate that has similar bulk properties to the API was selected based on a statistical analysis of the Rutgers material database. Feeding performance was evaluated to make sure the accurate delivery of material. For the continuous blender, process parameters such as blender speed and throughput were varied for blender characterization of six blends, which each contained different grades of excipients. The residence time distribution of the material in the continuous blender was performed using the pulse of a second surrogate, which was similar to the API and easily detectable by the spectroscopic methods and characterizing the profile of material as it exited the blender. By performing each of these steps, operators were able to identify the suitability and characteristics of blends based on the performances in the unit operations.

Methods: In order to study the process development, the formulation was identified containing 60% API surrogate, 34% excipients, 5% Crospovidone XL, and 1% Magnesium Stearate. Six different excipients including Prosolv 50, Prosolv 90, Avicel 101, Avicel 200, Pregel Starch and Lactose 310 were individually used to make six unique blends. Once the materials were shown to accurately feed, the materials were set up to perform blender characterization studies, including mass holdup of the blends, mean residence time, and the number of the blade passes the material undergoes as it passes through the blender. The values were studied by varying the critical process parameters of throughputs, blender speed and blend properties. Next, the residence time of the blender was characterized. This was performed by adding a pulse of a property-matched trace to an API surrogate at the entrance of the blender and using near-infrared spectroscopy (NIRS) to quantify the pulse as it exited the blender. For each process condition, two pulses each with a mass of 15g were separately introduced to the inlet of the blender when the system reaches the steady state.

Results: Confectionary Sugar was identified as the property-matched surrogate for the API. Feeding experiments were performed for the surrogate and API and similar behavior was observed. In Figure 1, the MRT was obtained at 10kg/hr and 25kg/hr for all six blends are shown as a function of the blender speed. The plot shows the MRT decreases with the rising of the blender RPM for all blends. At higher RPM, MRT appears to converge for six different blends, which means the impact of grades of excipients on the MRT is significant at relatively lower blender speed; however, at higher RPM the blender speed is the most dominant variable studied. In Figure 2, the results show the variation of blender speed and throughput versus the number of the blade passes. At 225 the blend would experience the most shear, which could either result in an optimal mixed system or a risk if the blend/constituents are shear-sensitive. With regard to the effect of the throughput, higher throughput led to lower blade passes as the material passes through the blended more quickly. The results of the predicted tracer concentration from the RTD experiment are shown in Figure 3. The distribution shows a good reproducibility for different replicates. The results show that higher peak intensity is observed with higher RPM while broader peaks are observed at lower RPM. Therefore, at higher blender speeds, a perturbation at the feeder level will pass through the unit operation with less filtering and propagate further down the manufacturing line.

Conclusion: Based on the study, we could understand the potential of the API manufacturability in a continuous line, including how critical process parameters and formulation development could lead to identifying optimal conditions for each. For a formulation with high drug loading, high throughputs with faster blender speed would result in less filtering of the material flow variations from the loss-in-weight feeders, due to low residence time and a decrease in shear strain to powder mixtures. However, at higher RPM, the residence time of each of the blends converge, even at different throughputs, indicating that at a high enough blender speed, the effect of the other variables are less significant.

Mean Residence Time (min) vs. blender RPM



#of blade passes vs blender RPM



Predicted Tracer Concentration